Abstract: Symbolic inference systems have been successfully used in the process control and me- dical diagnosis.But there are some problems that are hard to solve with pure symbolic processing te- chnology,such as vision function.Neural network-based systems are now used in these fields and show that there are hopes to solve these problems.Because the neural network learning efficiency is low and the time of training network is long in some cases,neural networks have not gained increasingly wide application.In this paper,the hybrid system,i.e.symbolic-neural network system,is presented. This system makes good use of the advantages of symbolic and neural network system,so it has both good thinking ability in terms of logic and good thinking ability in terms of images.